Does Transparency Lead to Pay Compression? Alexandre Mas * Princeton University and NBER February 2016 * Industrial Relations Section, Firestone Library, Princeton University, Princeton NJ 08544, [email protected]. I am grateful to Will Dobbie, Matthew Gentzkow, Ilyana Kuziemko, Emmanuel Saez, Jesse Shapiro, and Orie Shelef for helpful comments, as well as seminar participants at University of Chicago, Harvard, UC Berkeley, NBER, Princeton and Yale. Mingyu Chen, Kevin DeLuca, Kwabena Donkor, Helen Gao, Disa Hynsjo, Samsun Knight, Rebecca Sachs, Dan Van Deusen, Jessica Wagner, and Yining Zhu provided excellent research assistance. This project has been approved by Princeton’s Institutional Review Board.
75
Embed
Does Transparency Lead to Pay Compression? February 2016
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
compensation-data for the presidential memorandum, and https://www.sec.gov/news/pressrelease/2015-160.html for
the SEC rule of CEO pay relative to the median worker. 2 See, for example, The Economist, “Sunshine or colonoscopy?” November 19, 2011.
http://www.economist.com/node/21538774 3 Greenstone, Oyer, and Vissing-Jorgenson (2006) find evidence that mandated disclosure requirements in the 1964
Securities Act led managers to focus more on maximizing shareholder value. 4 See Di Tella and Fisman (2004), Diamond (2013) and Brueckner and Neumark (2014) for evidence of rent
extraction in the public sector.
4
Djankov et al. (2010) find that public disclosure of politicians’ income is associated with lower
perceived corruption and better government.
Transparency may also lead to pay compression if there is public sentiment against high
levels of compensation, even if compensation is in line with fundamentals. Such a response
would be consistent with inequality aversion on the part of the public (Fehr and Schmidt 1999).5
There has been speculation in the executive compensation literature that this type of “populist”
response to seemingly high levels of compensation has contributed to lower executive pay in
publicly traded companies where top salaries are disclosed (Jensen and Murphy 1990), though
there is little quantitative evidence of this phenomenon.
Understanding the effects of transparency on wage setting, as well as the underlying
mechanisms, is important for guiding policy. Estimating this effect is challenging, as it requires
finding variation in transparency at an organizational level, as well as data on wages. This paper
seeks to overcome these difficulties by examining how a 2010 California mandate that required
cities to disclose municipal salaries affected the compensation of the Chief Administrative
Officer (“city manager”) position—typically the highest paid city employee. The research design
exploits the fact that prior to the mandate a subset of cities (“previous disclosure” cities) had
already disclosed the salaries of their top managers.6 Using the Internet Wayback Machine and
archives of more than three hundred local newspapers, I identify cities where the salaries of city
managers were already disclosed to the press or on their websites at the time of the mandate.
Prior to the mandate, 63 percent of cities already disclosed the salary of the city manager. I
5 A related mechanism is morale considerations on the part of workers since transparency can lower job satisfaction
(Card et al. 2012) and being paid below expectations can lead to declines in productivity (Greenberg 1990; Krueger
and Mas 2004; Mas 2006; Mas 2008; Cohn et al. 2014). Employers might internalize these fairness concerns when
setting pay (Frank 1984, Akerlof and Yellen 1990, and Bartling and von Siemens 2010). Transparency might also
reduce gender and race wage gaps by making it easier to compare wages of workers in similar jobs; this was, in fact,
the stated motivation behind the 2014 presidential memorandum referenced above. 6 This strategy is similar to the one used by Bo, Slemrod, and Thoresen (2014) who study the effects of disclosure on
tax avoidance in Norway.
5
compare these cities to other cities where the mandate represented the first recorded disclosure of
city manager salary (“new disclosure” cities). I also make comparisons to wages in Arizona
cities, where there were no changes in disclosure policy.
One difficulty when studying the effects of disclosure policies is that pre-disclosure
information is not typically available. In order to examine pre-mandate trends, I made a public
records request to all 482 cities in California for 1999-2012 payroll records and contracts of city
managers, and to the California Public Employees' Retirement System (CalPERS) for 2001-2012
earnings records of employees who contributed to CalPERS pensions.
The evidence suggests that compensation is sensitive to increased transparency. Comparing
the evolution of wages in cities that previously did and did not disclose salaries, I find that salary
disclosure reduced compensation of city managers by an average of approximately 7 percent.
These cuts occurred both in cities where managers remained in their position, and in cities where
managers changed. Interestingly, given the evidence on firms’ reluctance to cut nominal wages
(Bewley 2012), these cuts were largely nominal. Wage cuts were substantially larger in cities
where compensation was initially higher, particularly cities where the city managers were paid
more than $200k annually prior to disclosure (the mean salary was $193k in 2009). There was no
relative decline in the 50th
, 75th
, and 90th
percentiles of the city wage distributions, on average,
implying that reductions at the top of the wage distribution reflect pay compression.
Importantly, this wage effect does not appear to be the result of citywide furloughs or
budget cuts following the 2007 recession. I find no evidence of differential changes between new
and previous disclosure cities in average earnings of municipal employees excluding the city
manager, number of municipal employees, or the average income levels of residents.
Additionally, the wage reductions came after furloughs peaked in California.
6
To assess whether these wage cuts were the result of greater accountability, I use the
estimated relationship between wages and city characteristics in cities that voluntarily disclosed
wages before the reform to predict wages in cities that did not disclose this information. I then
test whether the reduction in salaries in these cities is larger when there is a greater positive
residual wage, as would be the case if wages were driven up beyond market levels in secrecy. I
find that that this is not the case: wages fall as much in positive and negative residual cities on
average. This finding suggests that wage cuts were not the result of the discovery of managers
who exploited secrecy to inflate their wages, in general.
I also examine the effects of disclosure on manager turnover. I document that the policy
was associated with approximately a 75 percent increase in the quit rate. I find suggestive
evidence from a review of city manager biographies that, in turn, these cities experienced
increased difficulty attracting qualified managers to fill vacancies. Departing managers tended to
leave for other positions, either in the public or private sector, or retire and collect pensions.
Those who took other city manager positions had pay increases, on average.
Overall, the evidence is more consistent with the “populist” explanation of an aversion to
large salaries. Disclosure pressured cities to lower salaries, particularly large salaries, but not
salaries that were higher than predicted by fundamentals. The high levels of voluntary separation
following disclosure provide further support for this conclusion, since in these cases it was
evidently not possible to renegotiate compensation to transfer surplus to taxpayers. Providing
additional support for this conclusion, the new disclosure effect was more pronounced in cities
where voters had stronger redistributive preferences, as proxied by the Obama vote share in the
2008 election.
7
In addition to shedding light on the role of information on the wage structure, this paper
contributes to the literature on the causes and effects of government transparency (e.g. Djankov
et al. 2010) and to the related literature on the relationship between the press and political
accountability. The findings in this paper suggest that exposure to media had a significant effect
in restraining wages at the top of the wage distribution. This finding is consistent with Snyder
and Stromberg (2010) who find a relationship between press coverage and politicians’ actions
and policies.7
Section I. Pay Disclosure in California
In July 2010, an investigative report by the Los Angeles Times revealed that the city
manager of Bell, California (population 35,000) was being paid close to $800,000 annually
(Gottlieb and Vives 2010). While there were no legal limits on city manager compensation in
California (though in this case there turned out to be illegally written employment contracts), this
compensation was considered by many to be excessive. As a direct consequence of this scandal,
in August 2010 the State Controller John Chiang initiated the “Local Government Compensation
Reporting Program.” This program required salary information for elected officials and other
public employees to be clearly stated on city websites, and that the information be transmitted to
the State Controller’s Office and posted on its website (gcc.sco.ca.gov). The website went online
in October 2010 with almost universal compliance. By the end of 2012 the website had almost 6
million online views.8,9
7 On the role of the media and information on electoral outcomes and political accountability, see also Mondak
(1995), Besley, Burgess, and, Prat (2002), Arnold (2004), Reinikka and Svensson (2004), Besley and Pratt (2006),
Brunetti and Weder (2003), Gentzkow (2006), DellaVigna and Kaplan (2007), Ferraz and Finan (2008), Gentzkow,
Shapiro and Sinkinson (2011), and Pande (2011). 8 See http://www.acwa.com/news/state-legislation/state-controller’s-public-pay-website-gets-overhaul.
9 The Bell scandal led to increased attention and media interest in public sector salaries, and a number of newspapers
requested salaries of municipal employees from cities in their markets. As a result, the treatment is broader than the
State Controller mandate and includes all post-Bell media exposure.
8
A review of local newspapers around this time reveals cases where this disclosure led
residents to question city councils and management. For example, in the City of Lindsay, there
was a reported “outcry of shock and alarm” by the public over the city manager’s $214,405
salary, which many considered excessive.10
Disclosure led to “rumors that city water bills had
been raised to pay for abnormally high salaries.”11
Highlighting the opacity of the pay setting
process in this instance, a member of the city council stated that she “understood and shared
much of the sticker shock” because she knew “what the original contract with [the] city manager
had been, but not what his salary and benefits had risen to.”12
In other cases newly disclosed city
manager salaries were used as fodder in local political campaigns (Vorderbrueggen 2010) and
reported extensively on websites of watchdog groups.13
Following disclosure there were press
reports that city manager salaries were being driven down. A municipal recruiter was quoted
saying in 2011 (in the context of both the Bell scandal and the recession) that for “every city
manager search I’ve done since last year, the council is looking to pay the new city manager less
than the previous city manager.”14
With the goal of learning more about the public reaction to disclosure I spoke to city
council members at four cities, three new disclosure and one previous disclosure, about the
public response to increased transparency.15
A councilmember in a new disclosure city stated
that the topic of the city manager salary was a central topic of public discussion in city council
meetings immediately after the release of municipal salary data and while he believed that
10
Quoted in Kimball, Pam. 2011. “Lindsay Council Member Responds to Editorial,” Porterville Recorder,
(September 24). 11
Ibid 12
Ibid 13
Examples include the OC Reporter Watchdog site, many blogs focused on particular cities such as the “Lakewood
Accountability Action Group”, and the Transparent California website. 14
Quoted in Marois, Michael and James Nash. 2011. “California’s Top 10 Paid City Managers Raked in $4.7
Million” Bloomberg Business (June 13). 15
All spoke on the condition that they would not be identified in the paper.
9
compensation was set appropriately, disclosure and the resulting public response factored into
setting a low cost of living adjustment for the manager’s salary.16
In a second new disclosure city
a councilmember related that the city council was cognizant of public perceptions on
compensation when they were hiring a new city manager after disclosure and this made it harder
to find a qualified city manager.17
In a third new disclosure city a council member noted
increased media attention after disclosure, but did not believe that the council changed
compensation practices as a result of this attention.18
In the previous disclosure city a
councilmember related that there was increased public interest in the city manager’s salary after
the Los Angeles Times Bell report in 2010, but the city manager salary had been displayed on the
city website for many years and she felt that it was easy to respond to questions about pay as
they were no different than questions that they had received from the press in the past.19
Prior to the 2010 mandate there was one notable event relating to transparency. In August
2007 the California Supreme Court issued two rulings requiring disclosure of individual public
employee names, salaries and other employment information, but only when requested by the
public (Mintz 2007). Before this ruling, cities were not obligated to disclose compensation of
city employees by name to the public or to the press. The ruling led to several news outlets
obtaining and reporting compensation information, notably a consortium of San Francisco Bay
Area newspapers that published an online database in 2009 that included employee
compensation for fifty cities and other public entities.20
16
Telephone interview, October 7, 2015 17
Telephone interview, September 14, 2015. 18
Telephone interview, August 21, 2015. 19
Telephone interview, September 16, 2015. 20
As discussed in the Data Appendix, I will drop from the analysis cities for which first wage disclosure occurred in
2009, as there is some ambiguity about whether they are treated. The estimates are robust to inclusion of these cities
(see Online Appendix Table A1).
10
In principle, it would also be interesting to study the effects of the 2007 ruling; however, I
focus on the 2010 mandate because it does not appear that the 2007 ruling led to significant new
disclosure, and certainly not close to universal disclosure.21
Even the 2009 database mentioned
above led to new disclosure of city manager compensation for only ten cities. Figure 1 plots
mentions of “city manager” and “salary” in California by year as a share of all articles published
by California newspapers in NewsLibrary.com, an online newspaper archive.22
There is little
visual evidence of increased coverage of city manager salaries following the 2007 Supreme
Court ruling. There is a small increase in mentions in 2009, and a larger increase in 2010.23
Thus,
the distinction between passive disclosure (no obligation to post salaries) versus active disclosure
(obligation to post salaries) appears to be important.
One reason why the 2007 ruling engendered a limited response is likely that cities found
ways to delay fulfilling requests. For example, one of the Los Angeles Times reporters who broke
the Pulitzer Prize winning story on compensation in Bell, California describes the process by
which they obtained the information: "Literally every day, I'm calling the city clerk…I'm telling
her, `Listen, are we getting the documents? I really don't want to sue you, but we will, and when
we go to court, and we win, because we will, we'll ask the judge to make you pay our legal bills,
because that's what the [public records] statute says.’ The city manager, Robert Rizzo, finally
relented, but they had to meet him at a conference room near a city park for kids. That was weird
21
Using the data collection procedure described below, I find that 38 percent of cities had not disclosed city
manager salaries before 2010. 22 Specifically, for every year I search for keywords (“City Manager” OR “City Administrator” OR “Town
Manager”) in the first paragraph of the article and “Salary” anywhere else in the text. I divide the resulting number
of search results by the total number of articles in that year for the California press in the archive and then divide
this ratio by its value in 2000. Restricting the first set of key words to appear in the first paragraph reduces noise.
Taking a random sample of 40 search results in 2006 I found that the restricted search had 27 relevant search results
while the unrestricted search that allows the terms (“City Manager” OR “City Administrator” OR “Town Manager”)
to appear anywhere in the text had only 10 relevant results. 23 It is also possible that the disclosure of Bell salaries had an additional effect in turning the public’s attention
towards city compensation.
11
enough — but nine city officials and lawyers showed up.”24
Given these obstacles, the typical
resident (or even reporter without access to a legal department) would likely have had a difficult
time obtaining compensation information if city officials were inclined to prevent disclosure.
Section II. Municipal Governance and Compensation
Most California cities have a “Council-Manager” form of governance. Under this
arrangement, the city council, which is elected by voters, is responsible for setting broad policies.
The city council appoints a professional manager who is the head of administration. City
managers are typically in charge of day-to-day operations in the city, as well as developing a
budget, promoting economic development, collective bargaining, managing staff, and hiring.25
City managers also play an important role in generating tax revenues and providing public
services efficiently. In California, cities are constrained in their ability to raise revenues through
property tax increases due to California Proposition 13. Commercial development and resulting
sales tax revenue is one of the few ways that cities can raise revenues to make new investments,
and a city manager that can attract developers can be valuable (Lewis and Barbour 1999).
Mistakes by city management can have persistent negative consequences for cities, as seen in a
number of cases including Stockton, Vallejo and San Bernardino where financial
mismanagement resulted in costly bankruptcies (Winegarden 2014).
City managers are usually the highest paid municipal employees. Their compensation is
negotiated with the city council, often in closed session. Contract terms vary from setting pay
annually, to contracts that specify compensation over two or three year terms. Based on a
24 Quoted in Folkenflick, David. 2010. “How the L.A. Times Broke the Bell Corruption Story,” NPR.org
(September 24); available at http://www.npr.org/templates/story/story.php?storyId=130108851. 25 An alternative form of governance is Mayor-Council where an elected mayor serves as the city’s chief
administrative officer. These cities often have professional city managers that report directly to the mayor. See
Levin and Tadelis (2010) and Enikolopov (2012) for additional background on forms of city governance and the role
of the city manager.
12
random draw of ten contracts obtained through public records requests, the average term over
which compensation is pre-specified is two years. Pay raises can be left to the discretion of the
city council or can be linked to cost of living indices, compensation of other city employees (e.g.,
at least 10 percent above the next highest paid employee) or city managers in other cities. City
managers are at-will employees and can be fired at any time, though contracts may specify
severance payment depending on the term and circumstances of separation.
There is no legal limit to how much city managers in California can earn. City manager
compensation is driven at least in part by competitive forces, as shown by Enikolopov (2012)
who provides evidence that city manager compensation in the United States is sensitive to
manager performance. There may also exist institutional factors that decouple compensation
from market forces. Even when the city council has discretion to set a salary, their incentives
may not align with voters, particularly if salaries are not public. In the most extreme cases, the
city council and city managers may collude to boost each others’ compensation.26
Section III. Data
The data for this project comes from multiple sources, including public records act requests
and newly digitized archival documents. I describe them briefly here. Additional details on data
and sample selection are available in the Data Appendix.27
Compensation
City manager compensation for 2000-2012 was obtained from public records act requests
for payroll records and contracts of all 482 California cities.28
Since there was not universal
26 California has two types of cities: general law and charter cities. In general law the compensation of city council
is regulated while in charter cities council compensation is unregulated. City type is determined by referenda. 27
All data collection relied heavily from the help of the research assistants acknowledged above. Implicit in the
discussion below, research assistants did a lot of the arduous data collection work.
13
compliance to these requests, these records are supplemented with data obtained through a public
records act request to CalPERS for earnings histories of all municipal employees who were
employed during 2001-2012 and contributed to CalPERS pensions.29
Whenever possible I
construct longitudinal histories of city manager compensation using these sources. When there
are gaps, I use publicly available data in the State Controller website for years 2009-2012,
salaries found from Wayback Machine historical snapshots of city websites that reported city
manager salary, as well as salaries found in newspaper archives. This data collection effort
resulted in compensation histories for the city manager position for 76, 92 and 98 percent of
cities for 2001-2012, 2005-2012, and 2009-2012 respectively.30
I use Medicare earnings for 68
percent of cities for which I have data and base salaries for the remainder. The Arizona League
of Cities provided hardcopies of city manager salary histories for years 2004-2012 that I
digitized for this study. In what follows, city manager compensation will refer to the salary
compensation for the city manager position for a given city. For example, if there are different
managers in consecutive years, the change in compensation is the difference in their earnings. In
Section IV, I discuss whether and how changes to non-salary compensation may affect the
interpretation of my findings.
I construct measures of compensation for other municipal employees using CalPERS data
for years 2001-2012 and public data from the Local Government Compensation Reporting
Program for years 2009-2012. I compute average earnings (excluding the city manager), 50th
,
28 This public records request is only possible because of the Supreme Court ruling that city employee wages are in
the public domain. 29
CalPERS has earnings records for employees who contribute to pension benefits through this system. Enrollment
in CalPERS depends on city and occupation. In my data, in 2009 38 percent of California municipal employees
were enrolled in CalPERS and 90 percent of cities had at least one employee enrolled. 30
The reasons given for cities not providing complete information include record retention policy, old IT systems,
lack of staff, as well as non-response to my inquiry. For all requests I compensated the cities when necessary for the
costs of retrieving the records. Cities varied from not charging anything to charging upwards of $500 for the
information.
14
75th
, and 90th
percentile earnings of CalPERS enrolled employees over years 2001-2012.
Because CalPERS enrollees are only approximately 38 percent of all municipal employees, to
assess robustness I also construct these measures using the universe of municipal employees for
2009-2012 using the State Controller data, which includes one complete year of pre-mandate
data.
City Characteristics
Point-in-time city characteristics are five-year averages from the 2009 American
Community Survey and the 2007 Census of Governments. The controls utilized in the main
analysis are log population, log median housing values, log average household income, percent
of residents who are Hispanic, percent of residents who are black, percent of homes that are
renter occupied, percent of population (25+) with at most a high school degree, percent of
employed working in construction, and number of full-time equivalent city government
workers.31
I derive average resident income by year for the period 2004-2012 using annual zip-code
level data on resident income from the Statistics of Income program of the Internal Revenue
Service (SOI/IRS) that are then aggregated to the city level using zip code population weights.
Separation, City Council and Election Data
Data on manager turnover was obtained by digitizing the California Roster, a directory of
municipal elected officials and high-level managers published by the California Secretary of
State supplemented with online searches of city manager biographies since the roster is not
always up to date. These data span the years 2005-2013. To identify quits versus other reasons
for separation, for each record of separation I conducted an online search for press reports giving
background information on the reasons for the change. I code a separation as a quit if the
31
For all variables that are logged, I take the log of the five year average of the variable.
15
manager is reported to be leaving for another position, or if he or she is reported to be voluntarily
retiring (there is no mandatory retirement age).32
I read the articles to verify whether the manager
was voluntarily resigning versus being forced out, but it remains possible that some of the coded
resignations were actually involuntary.
I coded managers’ gender by comparing their first names to common female and male
names in the Social Security Administration names database. Where there was ambiguity, I
conducted an online search of the managers to identify their gender.
Data on City Council membership comes from digitized copies of the California Roster
for the years 2003-2012. Data on Obama vote share in the 2008 presidential election by city is
from the California Secretary of State website.
Disclosure Data
A key variable in this study is whether city salaries were in the public domain prior to the
2010 mandate. To obtain this information I searched for the city manager salary in pre-mandate
historical snapshots of each city’s website found on the Wayback Machine.33
If no salary was
found in the Human Resources, Administration or Finance directories of the website I recorded
the city as not posting the city manager salary online.
I also searched California newspapers over the period 2003-2009 for city manager salary
disclosure. The primary archive used is NewsLibrary.com, which at the time of search had
articles and transcripts for 338 California newspapers and TV stations. For every city in
California I searched for articles referencing the name of the city and city manager salary over
32
Retirement was determined by checking if the manager subsequently collected pension in the California Public
Employees' Retirement System, which is in the public record. 33 The Wayback Machine is a digital archive of websites (web.archive.org/). Specifically, I searched the last
snapshot of 2008.
16
January 1, 2003 to December 31, 2009.34
I coded a city as having prior city manager pay
disclosure if either it posted the city manager salary on its website or if the press reported city
manager salary sometime in the 2003-2009 period. I dropped the City of Bell from all of the
analyses so as to not confound the effect of disclosure with the Los Angeles Times report and
subsequent investigation into this city. I also excluded cities for which first disclosure was in
2009. Additional details on this data collection and sample selection criteria can be found in the
Online Data Appendix. These criteria result in 172 new disclosure cities and 296 previous
disclosure cities. A map of the location of these cities can be found in Figure 2.
Three appealing aspects of this definition are that a resident who wished to find city
manager salary could do so if the city had disclosed this information to the press at some point in
the recent past, it reflects the dichotomous nature of disclosure, and past disclosure to the press
likely signals that the city has a stance towards transparency. To the extent that I missed posted
salary information or if the information was disclosed by other means (not on websites or
newspapers), the operating assumption is that in such cases the information would have been
relatively more difficult to access than in cities coded as disclosing. My measure of disclosure is
therefore best thought of as an index that is related to previous transparency. If cities are
misclassified as not disclosing, this should lead to attenuation bias in the estimates. While the
focus of the analysis is on new disclosure based on this definition, I will show that there are also
wage effects, though smaller in magnitude, when the criterion for new disclosure is that the city
had not previously posted wages on their website.
Figure 3 shows the relationship between new disclosure and media coverage. As a measure
of media coverage I use the log of the total number of articles that mention the city between
34
This window was chosen to balance capturing news coverage over a recent period and feasibility, as reviewing
newspaper articles for mentions of city manager salaries is a time intensive process.
17
2003-2008. Note that any relationship found between new disclosure and this measure will not
be mechanical since the latter includes coverage on any topic (e.g. high school sports team
scores) and city manager/salary mentions are a miniscule share of all articles that refer to the city
(averaging 0.14 percent of all articles). The figure, which plots the probability of new disclosure
within bins separated by vingtiles of the media coverage measure, shows a strong negative
relationship between these two variables; new disclosure is much more likely in areas that had
little media coverage of any kind.35
Table 1 reports estimates from a linear probability model that predicts new disclosure using
media coverage and other city characteristics. In column (1) the explanatory variables are media
coverage, the percent of residents with no more than a high school degree, and log city
population. I include the latter two variables since exploratory analyses revealed that they are
highly predictive of whether there is new disclosure. The coefficient on the log total number of
articles is negative and significant. The coefficient implies that a 10 percent increase in prior
coverage is associated with a 0.76 percentage point decrease in the probability of new disclosure,
or a 2 percent decrease from a base of 37 percent. This estimate suggests that the disclosure
variable is partially related to media coverage in the city since cities with previous disclosure
have more press coverage on any topic, even conditional on population and other city
characteristics. The estimates on the other characteristics in column (1) imply that smaller cities
and cities with less educated residents are more likely to have new disclosure. While these
estimates should not be interpreted causally, the signs of the relationships are sensible. A larger
city will have more resources available to make information available as well as more interest
groups who demand it. Likewise, there might be more demand for salary information in places
35
This relationship is also observed when defining new disclosure only as the absence of information reported on
the city website (Appendix Figure A1).
18
where residents are more educated. In columns (2) and (3) I add additional characteristics but
conditional on the first three variables, these do not significantly predict new disclosure and their
inclusion does not change the relationships on the three variables in model (1).
Appendix Table A2 reports additional summary statistics organized by whether cities had
new disclosure. As already seen in Table 1, there are clear differences between these sets of
cities. These comparisons suggest that it is important to ensure robustness by controlling for a
rich set of city characteristics to verify whether the estimates pick up differential trends in
characteristics rather than the effects of disclosure. Table A2 also presents means of city
compensation. City managers in new disclosure cities earned 21 percent less in 2009 than
managers in previous disclosure cities, on average. This gap is largely accounted for by
differences in population and average income between these sets of cities.36
Both sets of
California cities have higher manager earnings than Arizona. Between 2009 and 2012 real city
manager compensation fell by 11.8 percent in new disclosure cities as compared to 4.9 percent in
previous disclosure California cities and 5.1 percent in Arizona cities. In the next section I will
estimate these changes more systematically, accounting for trends, city characteristics, and
regional shocks.
In column (1) of Table 2 I probe how disclosure affected news reports in these two sets of
cities. The column presents estimates of search results for mentions of city manager and salary
by new and previous disclosure cities, before and after the mandate period. For this search I used
a web scraper to count the number of search results with the above criteria, and I did not
manually verify whether the salary was actually reported.37
Therefore, there are some false
36 Controlling for log population and log mean household income alone reduces this gap to 3.7 percent. 37
The underlying data are city by year observations of search counts for (“YYY” AND ((“City Manager” OR “City
Administrator” OR “Town Manager” OR “Town Administrator”) AND (“Salary”)), where “YYY” is the city name
and the first two search terms are restricted to appear in the first paragraph of the article.
19
positives in the data using the search terms. I estimate a negative binomial model due to the low
counts and many zeros (particularly for new disclosure cities). I control for the log of the total
number of articles written about the city by year to ensure that any changes in counts are not
driven by a change in overall reporting. The estimates confirm that prior to 2010 new disclosure
cities have substantially fewer search results, with 64 percent (=(exp(-1.02)-1)*100) fewer
results in new disclosure cities relative to previous disclosure cities. This gap in the pre-mandate
period is mechanical based on how I constructed the new disclosure variable. The gap in search
results closes considerably, however, after the mandate in 2010, with the new disclosure gap
declining to 8.6 percent (=(exp(-1.02+0.93)-1)*100) fewer search results. This change, which is
not mechanical, is significant at conventional levels.
Figure 4 plots the estimated coefficients on the interactions of year dummies and a new
disclosure dummy from a negative binomial model. The figure shows a slight downward trend in
city manager/salary mentions prior to 2010 and then a discrete upward relative increase in city
manager/salary mentions in 2010 in new disclosure cities that persists through 2012. These
estimates confirm that the mandate led to considerably more press coverage on compensation for
cities that had previously not been mentioned in the press. However, they offer only a partial
view of how the mandate affected the diffusion of information as they do not account for the
people who accessed the salaries directly online.
Section IV. Disclosure and City Manager Outcomes
Earnings
Figure 5a presents visual evidence on the evolution of city manager salaries. The figure
presents nominal compensation of city managers in new disclosure and previous disclosure cities
for years 2001-2012. Specifically, for both sets of cities I regress log nominal city manager
20
compensation on year dummies and city fixed-effects.38
The figure plots the estimated
coefficients on the year dummies for new and previous disclosure cities normalizing 2009 to 0. I
also estimate and plot the coefficients from the fitted model for new disclosure cities where the
sample has been weighted to match a set of city characteristics (those listed in Section III) of the
previous disclosure sample. I use DiNardo, Fortin, and Lemieux (1996) (DFL) weights.
Prior to the mandate the growth rates of city manager compensation in previous and new
disclosure cities were close, with a slightly higher growth rate of compensation in new disclosure
cities in the mid-2000s. In 2010 nominal wage growth plateaued for both sets of cities, but in
2011-2012 nominal compensation declined sharply in new disclosure cities while for previous
disclosure cities wage growth remained stable in 2011, and then rose slightly in 2012. By 2012
the difference in compensation relative to 2009 was close to 7 percent. Reweighting the new
disclosure sample yields an almost identical pattern. Figure 5b plots the difference in the
(weighted) series with 95 percent confidence intervals. The difference in compensation relative
to 2009 in years prior to 2011 is never significant, but there are significant declines in
compensation in new disclosure relative to previous disclosure cities in 2011.39
The observed patterns in these figures are consistent with the mandate lowering salaries in
2011 and 2012. With respect to timing of the estimated effect, as previously discussed, city
manager contracts typically pre-specify compensation for a period of one or two years. We
would therefore expect to see the effect develop over the first two years following the mandate in
new disclosure cities as new contracts are negotiated. It is unclear whether we should expect to
see a divergence in compensation as early as 2010. While some cities may have reacted quickly,
38
City fixed-effects are necessary since the panel of cities is unbalanced. 39
A joint test of whether the coefficients on the New Disclosure terms are zero in all pre-mandate years yields a p-
value of 0.57.
21
it is likely that for most cities the changes would have taken more than a few months to affect
manager compensation.
Figure 6 plots both of the raw series against Arizona. One benefit of this comparison is
that Arizona was completely unaffected by the mandate, whereas previous disclosure cities in
California may still have been affected by the mandate and news coverage of Bell. Figure 6
shows that the growth rates of city manager compensation in Arizona and previous disclosure
cities in California line up well, both in the pre- and post- mandate periods.
I now turn towards estimating the average effect of disclosure in 2010, 2011, and 2012
relative to the pre-disclosure period. Table 3 reports estimates from variants of the following
log(Number of articles 2003-2008) -0.076 -0.074 -0.075
(0.020) (0.020) (0.021)
% of residents with at most a HS degree 0.006 0.006 0.006
(0.001) (0.002) (0.003)
log(population) -0.077 -0.081 -0.079
(0.017) (0.019) (0.020)
log(average HH income)
0.006 0.014
(0.090) (0.093)
% of residents who are black
0.003 0.003
(0.004) (0.005)
% of residents who are Hispanic
-0.000
(0.002)
R-squared 0.177 0.176 0.177
Observations 452 451 451
Notes: This table reports linear probability model estimates for the dependent variable of New
Disclosure which is 1 if the city did not have city manager salary reported in the press between 2003-
2008 or on its website in 2008. log(Number of articles 2003-2008) is the log of the number of articles in
the NewsLibrary.com database on any topic for years 2003-2008 for the city. Columns (2) and (3) have
one fewer observation than column (1) because log(average HH income) is missing for one city. The
unit of analysis is city. Robust standard errors in parentheses.
42
Table 2. Relation between New Disclosure and Time Varying City Characteristics
Negative
Binomial Model
OLS
City manager
mentions
ln(50th
percentile)
ln(75th
percentile)
ln(90th
percentile)
ln(Average
municipal
compensation)
ln(Num.
municipal
workers)
ln(Average
resident
income)
(1)
(2) (3) (4) (5) (6) (7)
New Disclosure * 2010 0.933
0.002 0.003 0.004 0.001 -0.002 -0.002
(0.258)
(0.013) (0.012) (0.013) (0.012) (0.019) (0.013)
New Disclosure * 2011 0.520
-0.005 0.002 0.002 0.002 -0.005 0.008
(0.198)
(0.015) (0.011) (0.012) (0.013) (0.020) (0.006)
New Disclosure * 2012 0.522
-0.008 -0.003 0.002 -0.014 0.013 0.005
(0.246)
(0.016) (0.011) (0.014) (0.017) (0.022) (0.009)
New Disclosure -1.019
(0.138)
log(Annual number of 0.553
articles mentioning city) (0.040)
City fixed-effects
X X X X X X
City characteristics * year
X X X X X X
R-squared
0.892 0.912 0.890 0.873 0.991 0.990
Observations 5436 4833 4833 4833 4833 4833 4043 Notes: Column (1) is a negative binomial model for years 2000-2012 where the dependent variable is the number of articles in NewsLibrary.com that contain terms "City
manager" and "Salary" by year and city. All dollar measures are nominal. 50th, 75th, 90th percentiles, average municipal compensation, and number of municipal workers are
derived from CalPERS earnings records for years 2001-2012 for 424 cities. Average resident income is derived from IRS/SOI zip code records for years 2004-2012 for 445
cities. City characteristics are log population, log median housing values, log average household income, percent of residents who are Hispanic, percent of residents who are
black, percent of homes that are renter occupied, percent of population (25+) with at most a high school degree, percent of employed working in construction, and number of
full-time equivalent city government workers. All main effects are included in the model, including year dummies. Standard errors clustered on city are in parentheses.
Observations 5108 5044 5044 5044 1783 2087 3351 Notes: All models estimated by OLS for years 2001-2012. The dependent variable in all models is log city manager
salary in 2012 dollars by city and year. Column (6) includes all Arizona cities and only new disclosure cities in
California. New disclosure is a city that did not previously have city manager salary online or reported in the press
(see text for details). City characteristics are log population, log median housing values, log average household
income, percent of residents who are Hispanic, percent of residents who are black, percent of homes that are renter
occupied, percent of population (25+) with at most a high school degree, percent of employed working in
construction, and number of full-time equivalent city government workers. Manager fixed-effects are a unique
fixed-effect for a manager in a city. All main effects are included in the model, including year dummies. Standard
errors clustered on city are in parentheses.
44
Table 4. Disclosure Effect by Type of Disclosure
(1)
(2)
New online disclosure* 2011-12 -0.040
(0.016)
Not online or in the news * 2011-12 -0.066
(0.017)
In the news but not online*2011-12 -0.006
(0.015)
Online but not in the news *2011-12 0.044
(0.045)
R-squared 0.898 0.899
Observations 5044 5044 Notes: Models estimated by OLS for years 2001-2012. The dependent
variable is log city manager salary in 2012 dollars by city and year. In
column (1) new disclosure is 1 if the city did not report the city manager
salary on its website. In column (2) new disclosure is further broken down
into three categories: “Not online or in the news” are cities that previously
did not have salary reports on their website or in the news, “In the news but
not online” are cities that had salary reports in newspapers but not online,
“Online but not in the news” are cities that had salary reports online but not
in newspapers, and the omitted group are cities that had salary reports both
online and in newspapers. “2011-12” is a dummy that is 1 for years 2011
and 2012. All specifications include city fixed effects and city
characteristics interacted by year dummies. See Table 3 notes for the list of
city characteristics. All main effects are included in the model, including
year dummies. Standard errors clustered on city are in parentheses.
45
Table 5. Heterogeneity by Pre-reform Wage Residual
(1) (2) (3)
1(Residual>0)*New Disclosure*1(2010) 0.025
-0.024
(0.028) (0.036)
1(Residual>0)*New Disclosure*1(2011) 0.006 0.005
(0.035) (0.042)
1(Residual>0)*New Disclosure*1(2012) 0.018
-0.016
(0.037) (0.048)
ln(2009 salary/200k≥1)* New
Disclosure*1(2010)
-0.198
(0.181)
-0.198
(0.182)
ln(2009 salary/200k≥1)* New
Disclosure*1(2011)
-0.664
(0.207)
-0.671
(0.204)
ln(2009 salary/200k≥1)* New
Disclosure*1(2012)
-0.657
(0.244)
-0.650
(0.237)
New Disclosure*1(2010) -0.015 0.041 0.054
(0.020) (0.021) (0.031)
New Disclosure*1(2011) -0.061 0.018 0.014
(0.026) (0.025) (0.041)
New Disclosure*1(2012) -0.077 0.014 0.023
(0.029) (0.033) (0.053)
R-squared 0.89 0.89 0.89
Observations 5083 5105 5083 Notes: Dependent variable is ln(city manager salary in 2012 dollars) by city and year for
2001-2012. "Residual" is the average residual computed using estimates from a regression of
log city manager salary on log city population, log average household income, and log
number of full-time municipal employee for each year in 2007-2009 in the previous
disclosure sample. The underlying regression used to estimate the residuals for 2009 is
shown in Appendix Table A6. ln(2009 salary/200k≥1) is ln(2009 salary/200k) if 2009
salary/200k≥1 and 0 otherwise. The models also include city fixed effects, interactions of
1(Residual>0) and year 2010-2012 indicators, interactions of ln(2009 salary), New
Disclosure and year 2010-2012 indicators, interactions of ln(2009 salary) and year 2010-
2012 indicators, and year dummies. All estimates are reported in Appendix Table A7.
Standard errors clustered on city are in parentheses.
46
Table 6. Disclosure and City Manager Turnover
(1) (2) (3) (4) (5) (6) (7) (8)
Sep. Sep. Sep. Sep. Quit Quit Quit Quit
New Disclosure*Post 0.215 0.231 0.231 0.138 0.178 0.172 0.152 0.076
To accurately identify the estimates for cities and towns in California and Arizona in the
ACS, I used the geographic data for places. Place data includes data for cities, towns, and census
designated places.
The demographic data come from the Selected Social Characteristics, Selected Economic
Characteristics, Selected Housing Characteristics, and the ACS Demographic and Housing
Characteristics files. Information on household types and educational attainment come from the
Selected Social Characteristics file. Information on employment status, industry, income, and
poverty level come from the Selected Economic Characteristics file. Information on housing
occupancy, value, and rent came from the Selected Housing Characteristics file. Information on
population, age, race, and ethnicity come from the ACS Demographic and Housing
Characteristics file.
Population and Housing Density Data
The population and housing density data come from the 2010 Decennial Census accessed
from the United State Census Bureau’s American FactFinder search tool, specifically CT-PH1
Population, Housing Unites, and Density: 2010 –State – Places. The dataset includes information
on population, housing units, area in square miles, and housing and population density per square
mile for every city and town in the United States.
Separation Data
Data on manager turnover was obtained by digitizing the California Roster, a directory of
municipal elected officials and high-level managers published by the California Secretary of
State, for years 2005-2012 supplemented with online searches of city manager biographies. The
online searches were necessary because the roster is not always up to date. To identify quits
4
versus other reasons for separation, for each record of separation I conducted an online search for
press reports giving background information on the reasons for the change. I code a separation as
a quit if the manager is reported to be leaving for another position, or if he or she is reported to
be voluntarily retiring.
I also use the city manager names to code manager gender for 2005-2012. I coded
gender based on a comparison of first name with common female and male names in the Social
Security Administration names database. Where there was ambiguity, I conducted an online
search of the managers to identify their gender.
Disclosure Data
To determine whether salaries were posted online I consulted historical snapshots of each
city’s website on the Wayback Machine (https://archive.org/web/) and investigated whether the
city manager’s salary was posted on the city’s website as of the end of 2008. Specifically, for
every California city, I searched the last available snapshot in 2008 in the Human Resources
Department, Finance Department, and the Administration directories of the city website
searching for the city manager salary. When available, this information is usually posted in the
Human Resources department directory. When no salary was found in any of these directories, I
recorded the city as not posting the city manager salary online.
I searched NewsLibrary.com for references to city manager salaries by city over the
period Jan 1, 2001 – December 31, 2009. Specifically, I searched for the name of the city in the
lead paragraph or title of the article and ((“City Manager” OR “City Administrator” OR “Town
Manager” OR “Town Administrator”) AND (“Salary”)) anywhere in the text. NewsLibrary.com
displays an excerpt of the first paragraph in the search results. I first verified whether the salary
was available there. If not, I assessed whether the article appeared relevant to the question and, if
5
so, I read the entire article to look for this information. The Los Angeles Times is not archived in
NewsLibrary.com and I did a separate search through their search engine using the same
methodology.
Dropped Cases
Starting with 482 cities in California, the following exclusions are made:
--14 cities where there is no record of a Chief Administrative Officer.
--The City of Bell because it was an investigation of this city that prompted disclosure, so as to
not confound an investigation with disclosure.
--14 cities where first disclosure occurred in 2009. This selection is imposed because these cities
are treated in that they first disclosed after the Supreme Court ruling, mostly through the Bay
Area online database, but the timing relative to the broader mandate is shifted by a year. The
estimates are unaffected by their inclusion as untreated or treated cities (Appendix Table A1).
--In specifications with controls, 8 cities with missing city characteristics
The main analysis utilizes 453 California cities without controls, 445 cities with all controls, and
85 Arizona cities.
6
Appendix Figure A1: Relationship between Media Coverage and Probability of New Disclosure; New Disclosure Defined as Absence of Salary Information on City Website
Notes: This figure plots the probability of new disclosure within twenty bins of the news coverage measure. The bin cut points are the vingtiles of the news coverage measure. The news coverage measure is the log of the total number of articles mentioning a city on any topic from 2003-2008.
.5.6
.7.8
.91
Prob
abilit
y of
new
dis
clos
ure
0 5 10 15 20News coverage vingtile
7
Appendix Figure A2. Share of all articles in the California Press referring to “Furloughs” (2000=1)
Notes: The data are from searches of California newspapers in NewsLibrary.com. The numerator is the number of articles referring to “Furloughs” in a given year and the denominator is the total number of articles in California. The share is expressed relative to 2000.
8
Appendix Table A1. Disclosure and City Manager Salaries; Including Cities for
New Disclosure * 2010 -0.003 -0.013 -0.011 -0.011 -0.011 -0.013 -0.023 (0.014) (0.015) (0.014) (0.013) (0.011) (0.030) (0.015) New Disclosure * 2011 -0.054 -0.067 -0.065 -0.065 -0.065 -0.084 -0.071 (0.017) (0.017) (0.018) (0.020) (0.018) (0.034) (0.020) New Disclosure * 2012 -0.064 -0.074 -0.072 -0.072 -0.072 -0.089 -0.076 (0.018) (0.016) (0.018) (0.023) (0.017) (0.036) (0.021) R-squared 0.892 0.899 0.917 0.943 0.953 0.950 0.947 Observations 5245 5164 5164 5164 1831 2185 3433 City fixed-effects X X X X X X X City characteristics * year X X X X X X County * year X City*linear trend X 2009-2012 sample X
Arizona Comparison X
Manager fixed-effects X
Notes: All models estimated by OLS for years 2001-2012. This table is identical to Table 3 in the main text but includes cities for which first disclosure was in 2009. The dependent variable in all models is log city manager salary in 2012 dollars by city and year. Column (6) includes all Arizona cities and only new disclosure cities in California. New disclosure is a city that did not previously have city manager salary online or reported in the press (see text for details). City characteristics are log population, log median housing values, log average household income, percent of residents who are Hispanic, percent of residents who are black, percent of homes that are renter occupied, percent of population (25+) with at most a high school degree, percent of employed working in construction, and number of full-time equivalent city government workers. Manager fixed-effects are a unique fixed-effect for a manager in a city. All main effects are included in the model, including year dummies. Standard errors clustered on city are in parentheses.
9
Appendix Table A2. Summary Statistics
(1) (2) (3)
(5) (6)
California New
Disclosure
California Previous
Disclosure Arizona
p-value (1)-(2)
p-value (1)-(3)
ln(population) 9.62 10.42 9.32
0.00 0.14
ln(Housing value) 12.81 13.10 11.98
0.00 0.00
ln(Mean Household Income) 10.90 11.09 10.71
0.00 0.00
ln(Government FTEs) 4.36 5.22 4.78
0.00 0.03
% Hispanic 39.89 30.24 31.03
0.00 0.01
% Black 3.588 3.798 2.07
0.70 0.01
% at most HS Degree 48.03 37.70 46.67
0.00 0.56
% Housing that is Renter Occupied 38.94 38.23 29.41
0.63 0.00
% Employed in Construction 7.348 7.124 9.60
0.63 0.00
% Employed in Retail 11.02 11.17 11.91
0.64 0.07
Female Manager (Yes = 1) 0.141 0.128
0.69
ln(2009 city manager compensation) 12.05 12.26 11.67
0.00 0.00
ln(2012 city manager comp.)-ln(2009 city manager comp.) -0.118 -0.049 -0.051
Female * New Disclosure -0.011 -0.003 -0.004 -0.030 0.002 -0.012
(0.024) (0.025) (0.028) (0.029) (0.025) (0.028)
Female 0.007 0.002 0.007 0.019 0.002 0.017
(0.014) (0.015) (0.016) (0.018) (0.015) (0.018)
City fixed-effects X X X X X X City characteristics*year X X X X X County*year
X City*linear trend
X
X City characteristics*year
X X *New Disclosure R-squared 0.93 0.92 0.94 0.96 0.93 0.96 Observations 3525 3475 3475 3475 3475 3475 Notes: Dependent variable is ln(city manager salary in 2012 dollars) by year and city for 2001-2012. The variable Female is an indicator for a female city manager. City characteristics are those used in Table 3 as well as an indicator for whether the city manager earned at least $200k in 2009. All main effects are included in the models.
11
Appendix Table A4. Disclosure Effect by Type of Disclosure
(1)
(2)
New online disclosure* 2010 -0.019
(0.016)
New online disclosure* 2011 -0.043
(0.018)
New online disclosure* 2012 -0.041
(0.020)
Not online or in the news * 2010 -0.018 (0.019) Not online or in the news * 2011 -0.065 (0.021) Not online or in the news * 2012 -0.070 (0.021) In the news but not online*2010
-0.017
(0.017) In the news but not online*2011 -0.010 (0.017) In the news but not online*2012 -0.006 (0.019) Online but not in the news * 2010
0.009
(0.041) Online but not in the news * 2011 0.047 (0.049) Online but not in the news * 2012 0.042 (0.053) R-squared 0.898 0.899 Observations 5044 5044 Notes: Models estimated by OLS for years 2001-2012. The dependent variable is log city manager salary in 2012 dollars by city and year. In column (1) new disclosure is 1 if the city did not report the city manager salary on its website. In column (2) new disclosure is further broken down into three categories: “Not online or in the news” are cities that previously did not have salary reports on their website or in the news, “In the news but not online” are cities that had salary reports in newspapers but not online, “Online but not in the news” are cities that had salary reports online but not in newspapers, and the omitted group are cities that had salary reports both online and in newspapers. All specifications include city fixed effects and city characteristics interacted by year dummies. See Table 3 notes for the list of city characteristics. All main effects are included in the model, including year dummies. Standard errors clustered on city are in parentheses.
12
Appendix Table A5. Relation between New Disclosure and Time Varying City Characteristics from Local Government Compensation Reporting Program Data
Observations 1775 1783 1783 1760 1695 Notes: 50th and 75th percentile city salaries, average municipal compensation, and number of municipal workers are derived from public data for the Local Government Compensation Reporting Program for 2009-2012. City characteristics are log population, log median housing values, log average household income, percent of residents who are Hispanic, percent of residents who are black, percent of homes that are renter occupied, percent of population (25+) with at most a high school degree, percent of employed working in construction, and number of full-time equivalent city government workers. All main effects are included in the model, including year dummies. Standard errors clustered on city are in parentheses.
13
Appendix Table A6. Estimates Underlying the Residual Calculation Used in Table 5 and
ln(Number of municipal workers) 0.078 0.065 0.062 (0.014) (0.016) (0.016)
Percent of population (25+)
-0.000 -0.001 with at most a HS degree (0.001) (0.001)
Percent of homes that are
0.002 0.006 renter occupied (0.001) (0.002)
ln(Housing Values)
-0.055 (0.061) ln(density) -0.054 (0.022) Percent of employment in construction 0.011
(0.004)
Observations 289 289 289
R-squared 0.54 0.54 0.57 Notes: This table shows the underlying estimates used to construct the residuals in Table 5 for year 2009 (the same models were estimated in 2007 and 2008. Dependent variable is log city manager salary. Sample is limited to previous disclosure cities in 2009.
14
Appendix Table A7. Additional Residual Heterogeneity Estimates and Models
City fixed-effects X X X X X X X City Characteristicsa X
X
City Characteristicsb
X X City Characteristicsc X X R-squared 0.89 0.89 0.89 0.89 0.89 0.89 0.89 Observations 5083 5105 5083 5083 5083 5083 5083 Notes: The table shows the full set of coefficients for the models reported in Table 5. See Table 5 notes for additional detail. This table also shows additional specifications using different characteristics to calculate residuals.
a Characteristics 1 are log city population, log average household income, and log number of full-time municipal employee.
b Characteristics 2 consists of Characteristics 1 plus percent of homes that are renter occupied and percent of population (25+) with at most a high school degree.
c Characteristics 3 consists of Characteristics 2 plus log median housing values, percent of employed residents working in construction, and log housing density.
16
Appendix Table A8. Relation between Magnitude of City Manager Wage Cuts and City Council Composition
City fixed-effects X X City characteristics * year X X Variable= Share of 2009 city council in office in 2005 X
Variable =Average tenure of city council in 2009 X R-squared 0.90 0.90 Observations 5030 5030 Notes: This table shows OLS estimates of the relationship between new disclosure, and interactions of new disclosure with measures of city council tenure, and log city manager compensation in 2012 dollars. The average tenure of the city council is based on truncated city council tenure for years 2003-2009 and a member of a city council who is observed in their position for all years 2003-2009 is assigned a tenure of six years. City characteristics are reported in Table 3.
17
Appendix Table A9. Relation between New Disclosure and Future City Council Composition
Share of city council in t-4 present in t
(1) (2)
New Disclosure * Post -0.009 -0.016 (0.046) (0.049)
Post 0.341
(0.027)
City fixed-effects X X
City characteristics*year
X
R-squared 0.627 0.638
Observations 901 888 Notes: This table shows OLS estimates of the effect of new disclosure on city council composition. The sample is California cities in 2008 and 2012. The post period refers to 2012. The dependent variable is the share of the city council in t (where t is either 2008 or 2012) that was present in t-4. City characteristics are those from Table 3. Robust standard errors in parentheses.
18
Appendix Table A10. Heterogeneity by 2008
Obama Vote Share
(1) (2) (3)
More than 50% Obama * 0.059 -0.016 0.002 New Disclosure *1(2010) (0.030) (0.027) (0.032)
More than 50% Obama * 0.006 -0.086 -0.046 New Disclosure *1(2011) (0.037) (0.039) (0.045)
More than 50% Obama * 0.009 -0.100 -0.099 New Disclosure *1(2012) (0.036) (0.046) (0.050)
New Disclosure*1(2010) -0.047 -0.003 (0.025) (0.021)
New Disclosure*1(2011) -0.070 -0.015 (0.032) (0.031)
New Disclosure*1(2012) -0.079 -0.014 (0.030) (0.036)
City fixed-effects X X X
City characteristics*year X X X City*linear trend
X X
City characteristics*year
X *New Disclosure
R-squared 0.90 0.94 0.95 Observations 5041 5041 5041 Notes: Dependent variable is ln(city manager salary in 2012 dollars) by year and city for 2001-2012. More than 50% Obama is an indicator for whether Barack Obama had more than 50% of the vote share in the city on the 2008 presidential election. City characteristics are those used in Table 3 and an indicator for whether the city manager earned more than $200k in 2009. All main effects are included in the models, including year dummies and the interactions of more than 50% Obama vote share with 1(2010)-1(2012).
19
Appendix Table B1. Disclosure and City Manager Salaries; No Imputed
(0.016) (0.018) (0.019) (0.025) (0.019) (0.037) (0.023) City fixed-effects X X X X X X X City characteristics * year X X X X X X County * year X City*linear trend X 2009-2012 sample X
Arizona Comparison X
Manager fixed-effects X
R-squared 0.893 0.894 0.918 0.937 0.952 0.950 0.946 Observations 4172 4122 4122 4122 1715 1842 2799 Notes: All models estimated by OLS for years 2001-2012. This table is identical to Table 3 in the main text but drops imputed values. The dependent variable in all models is log city manager salary in 2012 dollars by city and year. Column (6) includes all Arizona cities and only new disclosure cities in California. New disclosure is a city that did not previously have city manager salary online or reported in the press (see text for details). City characteristics are log population, log median housing values, log average household income, percent of residents who are Hispanic, percent of residents who are black, percent of homes that are renter occupied, percent of population (25+) with at most a high school degree, percent of employed working in construction, and number of full-time equivalent city government workers. Manager fixed-effects are a unique fixed-effect for a manager in a city. All main effects are included in the model, including year dummies. Standard errors clustered on city are in parentheses.